Learning gene networks underlying clinical phenotypes using SNP perturbation

Calvin McCarter, Judie A. Howrylak, Seyoung Kim

Abstract: Author summary We describe PerturbNet, a statistical framework for learning a gene network that modulates the influence of genetic variants on phenotypes, using genetic variants as naturally occurring perturbation of a biological system. PerturbNet directly models the cascade of perturbation from genetic variants to the gene network to the phenotype network, thus integrating the existing computational tools for eQTL mapping, GWAS, co-localization analysis of eQTL and GWAS variants, and gene network discovery under SNP perturbation within a single statistical framework. We demonstrate that PerturbNet improves statistical power for detecting disease-linked SNPs and uncovers gene networks mediating the SNP effects on traits, with computational efficiency that allows for human data analysis within several hours.